1 MASTER THESIS
What you see is what you get
The effect of image style and typing errors of online hotel reviews on purchase intention, helpfulness and trust
AUTHOR Annick Boer S2355701
EXAMINATION COMMITTEE Dr. R. S. Jacobs
Dr. J. Karreman
Faculty of Behavioural, Management and Social Sciences Master Communication Science
Specialization: Digital Marketing Communication
July 2021
<DATE>
2 Abstract
Aim: Because customers can only evaluate a hotel experience by visiting the hotel, customers
are dependent on online reviews to determine whether a hotel experience is worth purchasing.
As recent developments have allowed customers to, besides text, put images into their review, not only the way the review text is written, but also the kind of images that are included in the review can have an influence on the purchase decision process. Thus, the aim of this study is to investigate the effect of image style (in terms of professional images (studio aesthetics) versus amateur images (snapshots) versus no image), as well as the effect of typing errors in online hotel reviews on helpfulness, transaction trust, review trust and purchase intention.
Methods: In order to investigate the effect of image style and typing errors, a 3 (image style:
snapshot versus studio aesthetics versus no image) by 2 (typing errors: no errors versus errors) between-subjects experiment in the form of an online questionnaire was conducted, in which 183 people participated.
Results: The results of the experiment showed that typing errors had a negative effect on all
dependent variables. However, for image style, only an effect on purchase intention was found, as images with studio aesthetics showed significant higher purchase intention than snapshots and no image.
Discussion: Although the results on typing errors were in line with previous research and
showed the important role typing errors in online reviews play in the purchase decision process, the effects of image style did not meet the expectations. However, because it was showed that image style has an effect on purchase intention, it is essential to consider image style as an important determinant of buying behavior with regards to the hotel business.
Furthermore, practical recommendations and suggestions for future research are discussed.
Keywords: online hotel reviews, typing errors, image style, purchase intention, trust.
3 Table of contents
Abstract ... 1
1. Introduction ... 5
2. Theoretical background ... 9
2.1 Introduction ... 9
2.2 Trust ... 9
2.3 Helpfulness ... 12
2.4 Purchase intention ... 13
2.5 Typing errors ... 14
2.6 Image style ... 17
3. Method ... 23
3.1 Design ... 23
3.2 Stimuli and pre-test ... 23
3.3 Procedure ... 29
3.4 Participants ... 29
3.5 Measures ... 31
3.6 Validity and reliability of the measurements ... 32
4. Results ... 35
4.1 Manipulation checks ... 35
4.2 Hypothesis testing ... 37
4.3 Additional testing ... 40
5. Discussion ... 44
5.1 Discussion of the results ... 44
5.1.1 Typing errors ... 44
5.1.2 Image style ... 46
5.2 Theoretical implications ... 48
5.3 Practical implications ... 49
4
5.4 Limitations and future research ... 50
6. Conclusion ... 52
References ... 53
Appendix A ... 63
Appendix B ... 62
Appendix C ... 67
5 1. Introduction
Because of the emergence of technological advancements, the form of online reviews is changing. Recent developments have allowed consumers to provide more and richer
information when it comes to online reviews (Wu et al., 2020). Reviews have become more multimedia, as consumers are now able to not only review products through text, but also by uploading pictures, as well as videos (Xu et al., 2015). However, despite the fact that many review platforms, brand websites and social media platforms allow people to include images in their reviews, the influence of these images has only recently received some attention in academic research (Ma et al., 2018). Nonetheless, a study conducted by Zinko and colleagues (2021) showed the importance of images in reviews, as the results of the study concluded that reviews with images as well as text have a larger impact on trust and purchase intention than reviews without images.
The inclusion of images in online reviews is especially important for the hotel industry (Schuckert et al., 2015), because the customer cannot return a hotel experience or test it beforehand (Liu et al., 2020). As consumers develop a perspective on the hotel and the destination by looking at pictures (Trpkovski et al., 2018), visuals, as part of a review, are capable of creating a public image of a hotel, as well as display the experience other travelers had. Moreover, it can be said that images are of great importance to hoteliers, because
aesthetically pleasing images result in better customer recognition of the product (Trpkovski et al., 2018). Previous research has shown that images are the most significant factor to influence hotel booking decisions (Negri & Figolo, 2015).
An important difference between the hotel images that are available online, is the variation in style. Next to the professional photos that can be found for instance on a hotel’s website, lots of travelers also post their own pictures online. Although the style of these hotel images vary to a great extent, research on images has not investigated the impact of image
6 style on purchase intention, trust and helpfulness, as often only professional images were used in the research design (Zinko et al., 2021). Yet, this is of great importance, as consumer generated content varies in style to a large extent compared to professional images. Therefore, this study investigates image style by means of making a distinction between professional images that portray studio aesthetics and snapshots that are taken by customers of the hotel.
Next to the image, various characteristics of the review text also play a vital role in the purchase decision process. One of the characteristics that has gained quite some attention in academic research is whether the review contains typing errors or not (Risselada et al., 2018;
Cox et al., 2017; Cooper et al., 2020). Previous research has shown that typing errors
negatively influence the perceived helpfulness of the review, as well as the trust in the review an ultimately the purchase behavior of the customer (Schindler & Bickart, 2012). One of the reasons for this is that a text that contains typing errors is essentially harder to process, as the readability of such a text is much lower, making the review less helpful (Risselada et al., 2018). Moreover, the author of the review is seen as less conscientious when the text contains typing errors, which reduces the trust in the review (Cox et al., 2017). However, previous research has not taken into account the interplay between typing errors and image style. This gap should be addressed, as it gives more insights into which (combinations of) characteristics of online reviews have an influence on people’s buying behavior, as well as their purchasing experience. Furthermore, by taking into account both image and text, this study will better reflect on real life online review experiences (Zinko et al., 2021).
Thus, this study researches the effect of image style on purchase intention, trust and helpfulness by comparing the effect of a review without an image with the effect of a review with an image which portrays studio aesthetics and a review with a snapshot image.
Furthermore, the effect of typing errors in reviews on purchase intention, trust and helpfulness
7 is researched by comparing the effect of reviews with typing errors with reviews without typing errors. The following research question was therefore developed:
RQ: To what extent do typing errors and image style of an online hotel review have an effect on the experience of trust, helpfulness and purchase intention?
By answering this research question, this study contributes to and extents the already existing research on online reviews and will therefore advance the paradigm. Moreover, the results of this study are highly valuable for hotel businesses, as well as for booking and review websites, as it will help them to determine which characteristics of an online review positively effect the purchase behavior of their customers. This, in turn, can be used to, for example, select certain reviews on the basis of their characteristics and give them a more prominent position on the website page. This is also beneficial for the customer, as the results of the study can be used to make reviews more helpful and thus contain more valuable
information.
In order to investigate the effect of image style and typing errors, a 3 (image style:
snapshot versus studio aesthetics versus no image) by 2 (typing errors: no errors versus errors) between-subjects experiment was conducted in the form of an online questionnaire. The focus in this study is on positive reviews, as previous research has shown that consumers are less likely to post pictures with their reviews if they had a negative hotel experience (Trpkovski et al., 2018). Moreover, it has been found that positive online hotel reviews increase trust and purchase intention (Kusumasondjaja et al., 2012; Sparks & Browning, 2011). Using positive reviews will therefore reveal whether image style and typing errors will increase or decrease trust, purchase intention and helpfulness.
The remaining parts of this paper will elaborate on the conceptual framework and will develop hypotheses. Furthermore, a chapter will be dedicated to explaining the methodology
8 that is used in order the test these hypotheses. The results of the study will then be showcased, which will be followed by a chapter which discusses these results and a conclusion.
9 2. Theoretical background
2.1 Introduction
This chapter will define and elaborate on the concepts that are of interest in this study and will therefore develop hypotheses. Firstly, the concepts transaction trust, review trust, helpfulness and purchase intention will be discussed, as well as previous research that has been done on these topics with regards to online reviews. These concepts have played an important role in research on reviews, as they have a large effect on the purchase decision process of customers (Alfina et al., 2014). Therefore, these concepts are also considered to be important outcomes when investigating the effect of typing errors and image style. Secondly, typing errors and image style will be defined and this part will also touch upon various studies related to both concepts. Throughout the chapter, the hypotheses that will be tested are stated. Lastly, at the end of this chapter, a conceptual model will be presented.
2.2 Trust
Trust can be defined as “a party’s willingness to accept vulnerability, but with an expectation of confidence that the other party can be relied upon to not take advantage of the trustor”
(Bart et al., 2005, p. 134). The concept of trust has often been used as an outcome in research about online reviews (Zinko et al., 2019). The reason for this is that trust has a positive effect on purchase intention (Alfina et al., 2014). It has even been argued that trust is the most important indicator of purchase behavior, because trust indicates the expectation that an organization or service is dependable (Sparks & Browning, 2011). A lack of trust, therefore, can be seen as a barrier, which keeps people from making a purchase online (Wang &
Emurian, 2005). This means that when there is no confidence in a person or an organization, the likelihood of a purchase taking place is very small (Wang & Emurian, 2005). Online reviews are thus crucial for the development of trust, as the experiences other customers had can build on the belief whether a person or an organization is trustworthy or not (Sparks &
10 Browning, 2011). A study done by Tran and Strutton (2020) has for example shown that e- WOM positively influences trust and that trust, in turn, positively impacts customer loyalty.
Moreover, the establishment of trust is especially important for the hospitality
industry, as well as other experience goods, in order to reduce uncertainty and risk perception among customers (Cheng et al., 2019). The distinction is often made between search and experience goods, with search goods being a product for which a customer can retrieve information on its quality before purchasing it, such as a computer, and an experience good being a product where one has to purchase it in order to determine its quality, such as a hotel experience (Lee & Choeh, 2016). Because customers can only evaluate a hotel experience by actually visiting the hotel, they are dependent on other informational cues to determine whether a hotel experience is worth purchasing (Cheng et al., 2019). Reviews, as a kind of informational cue, are especially important, because they give customers access to prior service experiences, thus reducing risk and uncertainty and increasing trust in the hotel (Sparks & Browning, 2011).
Whether consumers think that the review actually contains a real prior experience has been found to have a great impact on the establishment of trust. The results of the study by Kim and Kim (2020) for example show that perceived authenticity of online reviews has a significant impact on the development of trust for customers. According to Kim and Kim (2020), the reason for this is that “travelers tend to use the reviews as a cue for assessment not of the users who posted but of the websites that host the users and provide this user-generated information” (p. 772). The extent to which a review is found authentic is therefore used by customers to judge whether a site or service can be trusted.
Previous research on trust in the context of online reviews has discussed different dimensions of trust (Sparks & Browning, 2011; Cheng et al., 2019; Kim & Kim, 2020;
Kusumasondjaja et al, 2012; Zinko et al., 2021). Zinko et al. (2021) for example describe the
11 dimensions transaction trust and interpersonal trust. In the context of reviews, interpersonal trust (also termed party trust (Tan & Thoen, 2000)) describes the trust one has in the person who wrote the review. However, because this study does not focus on the characteristics of the reviewer, interpersonal trust is not within the scope of this research. This study will focus on the other dimension that is discussed by Zinko et al. (2021): transaction trust. Transaction trust is described as “a mental state that determines whether the focal individual has sufficient trust to engage with in a transaction” (Zinko et al., 2021, p. 86). This therefore portrays the trust a customer has to experience in order to commit to a transaction and purchase the hotel experience. Customers who believe that the hotel will deliver on its promises, experience more trust to engage in a transaction with the hotel (Wang & Emurian, 2005). The reason that this dimension is chosen as a focus of this study, is that, logically, transaction trust is essential for customers to eventually purchase a hotel experience, as a customer will not purchase a hotel experience when they do not experience sufficient trust to engage in a transaction with the hotel (Zinko et al., 2021).
The other dimension that is of interest in this study, is described by Ahmad and Guzmán (2020) as message trust. This concept focusses on the extent to which a message is found to be trustworthy. In the context of online hotel reviews, the written review as well as the possibly attached image can be seen as the message that has to be trusted by the customer.
In the context of this study, we will continue using the term review trust. Grewal et al. (1994) for instance found that when a message is found to be trustworthy, the perception of product performance risk is lower, which therefore increases purchase intention. By including two different dimensions of trust in the study, we want to see whether differences in image style and typing errors of an online hotel review have a different effect on the trust people have in engaging in a transaction with the hotel and the trust people have in the review itself.
12 2.3 Helpfulness
Next to trust, review helpfulness has also been a largely studied outcome in the research on online reviews. A helpful review can be defined as “a peer-generated product evaluation that facilitates the consumer’s purchase decision process” (Mudambi & Schuff, 2010, p. 186).
Review helpfulness has been used as a way to measure how consumers evaluate a review (Mudambi & Schuff, 2010). Many retailers, like for example Amazon.com, already measure this by asking consumers whether they found a review helpful (Cao et al., 2011). As the amount of reviews on some products can cause information overload, the helpfulness voting that many retail sites nowadays offer can facilitate the decision process of consumers, as they see the most helpful reviews first. Therefore, making it easier for consumers to make a purchase decision (Cao et al., 2011).
What is important to note is that the determinants of a helpful review differ for various product categories. As was discussed previously, the difference here is made between search products and experience products. Research has found that the amount of reviews is more important for review helpfulness of experience products, whereas reviewer reputation and extremity has a bigger influence on the perceived helpfulness of reviews of search products (Lee & Choeh, 2016). According to Lee and Choeh (2016), the total number of reviews that a product has represents the amount of information that is available, meaning that a more popular product has more reviews. Because the sellers information that is available on experience goods is often not sufficient to make a purchase decision, the total number of reviews is used in the first stage of the purchase decision process and is used as a cue to determine whether a product is good or not (Lee and Choeh, 2016). So, although many reviews on a product can cause information overload, as previously stated, a large amount of reviews is also a sign of popularity for experience goods and thus probably good quality.
13 The study done by Mudambi and Schuff (2010) builds on the above mentioned
findings from Lee and Choeh (2016), as they found that extremity in reviews is less helpful for experience goods, like hotels. Moreover, the results of their study suggested that research depth is important for the helpfulness of a review, but less for experience goods than for search goods. According to Mudambi and Schuff (2010), the reason for this is that the product type influences the type of information that a customer is looking for. As it is much easier to give a textual description of the quality of a search good than of an experience good, people found long reviews for search goods much more helpful than for experience goods.
Furthermore, subjective information is used when deciding to purchase an experience good (as opposed to objective information when purchasing a search good), meaning that less detail is required and thus a shorter review is sufficient (Lee & Choeh, 2016).
2.4 Purchase intention
Purchase intention is a prominent concept in the literature about online reviews (Zhang et al., 2014; Jiménez & Mendoza, 2013) and can be defined as “a consumer's willingness to buy a given product at a specific time or in a specific situation” (Lu et al., 2014, p. 261). Although purchase intention does not directly translate to actual purchase behavior, it still remains important to study the intention to purchase. According to the theory of planned behavior (TPB), intention is a crucial intervening stage between one’s attitude and actual behavior (Hassan et al., 2016). In general, it can be said that the stronger the intention is to perform a certain behavior, the more likely it is that one actually executes the behavior (Ajzen, 1991).
Despite the fact that intention is not equal to behavior, many studies regarding many types of behaviors have proven the TPB to be reliable, and thus were able to predict behavior through intentions, to some extent (Ajzen, 1991; Gass & Seiter, 2015). In the context of online purchase intentions, a study done by Gu and Wu (2019) has found that intention to purchase was indeed a reliable measure for predicting actual purchasing behavior. Therefore, the
14 concept of purchase intention will be used in this study as a predictor for actual purchase behavior.
Research has shown that purchase intention is influenced by a variety of different elements (Zinko et al., 2019). Next to brand loyalty, general attitude and word-of-mouth, risk perception plays an important role in the formation of purchase intention. When people do not have sufficient information about a product, which is especially the case for experience
products, risk perception is rather high. But by reading online reviews, risk perception can be reduced, thus increasing purchase intention (Zinko et al., 2019). This is closely related to both transaction trust and review trust, as one can argue that when risk perception is reduced, both kinds of trust are increasing. As was mentioned above, previous research has confirmed this link between purchase intention and trust, as it has been found that trust positively influences purchase intention (Alfina et al., 2014). We therefore adopt the following hypotheses:
H1: Transaction trust positively influences purchase intention.
H2: Review trust positively influences purchase intention.
Just like trust, helpfulness has also been found to have an influence on purchase intention. A study done by Filieri et al. (2018) investigated the determinants of review helpfulness and its effect on purchase intention. The outcome of their study did not show a positive relationship between trust and helpfulness, but the study did find that helpfulness has a positive influence on purchase intention. Thus, we hypothesize that:
H3: Review helpfulness positively influences purchase intention.
2.5 Typing errors
Previous research has shown that various characteristics of the text of an online review play an important role in the purchase decision process of consumers. One of the ways in which text has an influence, is by making a review more (or less) helpful. Baek et al. (2012) for example studied review helpfulness in the context of dual process theories and focused on
15 which peripheral and central cues had an influence on the perceived helpfulness of online reviews. According to dual process theories, people process information in two types of ways.
When one uses heuristic information processing, they make use of the peripheral route, meaning that they make decisions based on associations they have with certain cues.
However, when one uses systematic information processing, they make use of the central route, meaning that they carefully consider all the information that is at hand. Taking the central route therefore takes much more cognitive effort and is much more extensive than the peripheral route (Baek et al., 2012). The results of the study show that when customers are in the stage of evaluation of alternatives, when shopping for a product, they tend to focus on central cues, meaning that when a customer is making a final choice, the content of the review is most important for the helpfulness of the review.
A research done by Korfiates et al. (2012) took into account other review
characteristics, and suggested that the word count of a review is not the best predictor of the perceived helpfulness of a review. The results showed that readability had a greater effect on review helpfulness than review length (Korfiates et al., 2012). This was also supported by Li et al. (2020), as they found that reviews with a short length and high readability could achieve the best performance and that overall high text quality could better serve customers and their decision making process. They noted that high readability and a short length of the review helped customers to accurately identify the reviewer’s opinion, which aids their purchase decision (Li et al., 2020).
A study done by Cao et al. (2011) also investigated the factors that influence the voting of helpfulness on online reviews with regards to text quality. They took into account basic characteristics (i.e. rating of the product, posting time), stylistic characteristics (i.e.
sentence length, number of long words) and semantic features (i.e. the meaning of words).
The results showed that all characteristics had a significant influence on helpfulness votes.
16 As was showcased in the paragraphs above, various characteristics of text have an influence on the helpfulness and thus the purchase intention of the customer (Mudambi &
Schuff, 2010; Baek et al., 2012; Korfiates et al., 2012; Li et al., 2020; Cao et al., 2011).
However, one characteristic that has also played an important role in the research on the effect of the review text is typing errors. Research has found evidence that typing errors have an effect on the purchase intention of customers. A study done by Akhtar et al. (2020), which was also done in the context of online hotel reviews, considered typing errors in the light of language expectancy theory. Language expectancy theory assumes that language is rule-based and because people create norms for appropriate language use, behavior change occurs when these norms, and thus the expectancy, is violated (Akhtar et al., 2020). As typing errors are a violation of language rules, they have a negative effect on various outcomes. This was also supported by the results of the study by Akhtar et al. (2020), which showed that typing errors indeed have a negative effect on purchase intention. We therefore hypothesize that:
H4: Typing errors in reviews result in lower purchase intention than no typing errors.
Next to purchase intention, trust has also played a role in research on the effect of typing errors in reviews. Cox et al. (2017) for example focused on the effects that typing errors had on perceived credibility of the review and made a distinction between
typographical (i.e. mechanical errors, such as mistyping) and orthographical errors (i.e.
cognitive errors, misspellings). The results showed that the effect of textual errors on credibility rely on general trust and that consumers with high trust negatively viewed typographical errors, as it showed carelessness, as opposed to orthographical errors, which indicate cognitive ability (Cox et al., 2017). Cox et al. (2017) argue that typing errors indicate a lack of conscientiousness, which is correlated with dishonest behavior. Therefore, a review with typing errors is seen as dishonest and is trusted less than a review that does not contain typing errors. Cooper et al. (2020) had similar results in that they found that typing errors in
17 online reviews have a negative impact on organizational attraction. The results of the research showed that both negative and positive reviews have a more positive impact when they do not include any errors. As the above mentioned research shows an effect of trust in general, we hypothesize that typing errors have an effect on both transaction trust and review trust:
H5: Typing errors in reviews result in lower transaction trust than no typing errors.
H6: Typing errors in reviews result in lower review trust than no typing errors.
Next to trust and purchase intention, typing errors also have an effect on the perceived helpfulness of a review. According to Risselada et al. (2018) content presentation
characteristics, such as typing errors, are important drivers for the perceived helpfulness of online reviews. The reason for this is that typing errors influence the way customers process the information in the reviews, and thus impact the helpfulness of a review. When a review contains typing errors, the readability of the review is much lower, which makes information processing less fluent (Risselada et al., 2018). The study by Risselada et al. (2018) therefore also concluded that typing errors had a negative effect on helpfulness votes. This is in line with the research by Schindler and Bickart (2012), as the results of their study show that
“stylistic elements (such as typing, spelling and grammatical errors) were associated with less valuable reviews” (Schindler & Bickart, 2012, p. 234). Therefore, the following hypothesis was constructed:
H7: Typing errors in reviews result in lower helpfulness than no typing errors.
2.6 Image style
Images have proven to be very effective in that they have the ability to catch the attention of a consumer much faster than only text would (Teo et al., 2019). Moreover, images are
processed faster than text and are capable of transferring much more information (Zinko et al., 2021). In order to explain how images are able to create more customer engagement than only text can, media richness theory (MRT) can be used (Zinko et al., 2021). According to Daft
18 and Lengel (1986), communications are considered rich when they “can overcome different frames of reference or clarify ambiguous issues to change understanding in a timely manner”
(p. 560). They therewith argue that some media are more effective when transmitting information, as they are able to transmit more context and cues. Because images are able to transmit much more information than text only, images allow reviewers to enrich their review and give a more complete picture of their experience, thus increasing engagement with the reviews (Zinko et al., 2021).
Image style can be defined in various ways. One way in which one can differentiate between image styles is by looking at image quality. When informally defining image quality, Ke et al. (2006) make a difference between professional photos and snapshots, stating that:
“we define professional photos as those that would be framed and hung on a wall, and snapshots as those that would stay in a photo album” (p. 419). However, another, more general, definition of image quality is given by Engeldrum (2004): “Image quality is the integrated perception of the overall degree of excellence of an image” (p. 448). In practice, this means that image quality, just as text quality, has various dimensions. According to Ke et al. (2006), a high quality images possess various characteristics. First, high quality images show simplicity in the sense that it is obvious what one should be looking at. This is often not the case for snapshots, as they are often busy and full with clutter. High quality images therefore use color and lighting contrast to let the subject pop out.
Another way in which high quality images differ from snapshots is realism. High quality images can be seen as surreal, because they use much brighter colors and place the subject in unusual settings. This means that high quality images often portray a scene that does not look natural, as one would not encounter it in real life. One might therefore argue that professionally taken pictures are surreal in the sense that they enhance reality.
19 Lastly, a difference that Ke et al. (2006) found between professional high quality images and snapshots is that overall basic techniques of high quality images are better.
Snapshots are much more often blurry, which is a result of poor technique and low quality equipment. However, it is important to note that blurriness and low quality equipment are not representative characteristics of low quality images anymore, as the devices that are used to take pictures have become much more advanced (Tprkovski et al., 2018). Thus, making it unlikely to snap a shot that shows blurriness.
Trpkovski et al. (2018) also addressed several visual features that were used to assess the image quality of online travel pictures. Just as Ke et al. (2006), Trpkovski et al. (2018) assessed image quality by means of brightness, colorfulness and contrast. However, Trpkovski et al also took into account noisiness and sharpness. They describe noisiness as randomly appearing dots in the photo, which is a result of random pixel variation. The results of their analysis of over 10.000 travel images showed that professional pictures of the hotel (which are for example pictures that are posted on the original hotel website) indeed have higher brightness, saturation, contrast and sharpness and less noisiness, thus resulting in higher quality images than those of travelers.
This study will define image style by making a difference between snapshots and studio aesthetics. Snapshots in this sense are pictures taken by customers of hotels, which portray clutter, such as personal belongings, and low contrast and brightness. Images which portray studio aesthetics, on the other hand, show simplicity, high contrast and brightness and are very colorful. These images are taken by a professional and in some sense can be seen as surreal, as the image of the hotel room displays an enhanced version of reality (Ke et al., 2006).
Because previous research has shown that images of high quality (such as studio aesthetic images) result in more positive affective experiences (Colliander & Marder, 2018),
20 one might argue that images with studio aesthetics result in more engagement than snapshots, and might therefore also result in higher transaction trust, review trust, helpfulness and purchase intention. Teo et al. (2019) validated this, as the results of their study on Instagram marketing show that high image quality positively influenced perceived product quality and purchase intention. This might therefore also be the case for online reviews with images that show studio aesthetics. However, because images in online reviews are often user-generated and therefore informal in nature, studio aesthetics might also reduce trust. According to Colliander and Marder (2018), snapshots therefore “hold greater congruence with the custom of the medium” (p. 35). Because reviews are written by laypeople, it would be expected that the reviews would not include high quality studio aesthetic images. A review containing a studio aesthetic image could therefore reduce the trustworthiness of the review. Previous research confirmed this as well, as it was found that snapshot aesthetics on a brand’s
Instagram account resulted in higher product recommendation and brand attitudes, as opposed to studio aesthetics (Colliander & Marder, 2018). When taking into account previous research, it can be said that studio aesthetics might disturb trust, helpfulness and purchase intention, as it reduces credibility of the review. Therefore, snapshots are more realistic and can give the customer a better view of the actual hotel quality.
Furthermore, as previous research and MRT have shown that images are able to transfer more information than text, it is believed that including an image in the review has a positive effect on the perceived helpfulness, transaction trust, review trust and purchase intention than a review without an image. This was also shown in the results of the study done by Zinko et al. (2021), as their study showed that reviews with images as well as text have a larger impact on trust and purchase intention than reviews without images. Thus, the
following hypotheses have been constructed:
H8: Snapshots result in higher purchase intention than studio aesthetics and no image.
21 H9: Snapshots result in higher transaction trust than studio aesthetics and no image.
H10: Snapshots result in higher review trust than studio aesthetics and no image.
H11: Snapshots result in higher helpfulness than studio aesthetics and no image.
Furthermore, it is believed that images can decrease the negative effect that reviews with typing errors have on helpfulness, transaction trust, review trust and purchase intention.
As this section showcased, images are more easily processed and are able to transfer more information than text only (Zinko et al., 2021). Because reviews with typing errors result in worse processing of information (Risselada et al., 2018), the inclusion of images to reviews with typing errors can decrease the negative effect that reviews with typing errors have on helpfulness, transaction trust, review trust and purchase intention in comparison to reviews which do not include an image (see figure 1):
H12a: The inclusion of an image in the online review decreases the negative effect of a review with typing errors on transaction trust.
H12b: The inclusion of an image in the online review decreases the negative effect of a review with typing errors on review trust.
H12c: The inclusion of an image in the online review decreases the negative effect of a review with typing errors on purchase intention.
H12d: The inclusion of an image in the online review decreases the negative effect of a review with typing errors on helpfulness.
22
Figure 1: Conceptual framework
The following chapter will touch upon the method that was used to test the hypotheses, as well as the participants that were included in the study. Furthermore, after the methodology is discussed, the results of the study will be showcased. These results will then be discussed in the discussion chapter, followed by a brief conclusion.
23 3. Method
3.1 Design
In order to examine the effect of image style and typing errors on purchase intention, helpfulness, transaction trust and review trust, an online experiment was conducted. The experiment was done in the form of a 3 (image style: snapshot versus studio aesthetics versus no image) by 2 (typing errors: no errors versus errors) between-subjects experiment. As can be seen in table 1 below, the experiment contains six different conditions, with one review per condition. The questionnaire was created with Qualtrics. In order to test the hypotheses, participants were asked to read and look at mock-up reviews. The reviews were written in the context of a hotel booking site. Therefore, the photos and texts that were included in the reviews concern a hotel experience.
3.2 Stimuli and pre-test
In order to investigate the effect of image style and typing errors on purchase intention, helpfulness, transaction trust and review trust, the mock-up reviews had to be manipulated in two different ways: in terms of text and in terms of image. The following section will discuss
Table 1 Conditions
Condition Typing errors Image style
1 Yes Studio aesthetic
2 No Studio aesthetic
3 Yes Snapshots
4 No Snapshots
5 Yes No image
6 No No image
24 how the mock-up reviews were manipulated and how a pre-test was done in order to test the extent to which people recognize the manipulations.
As stated previously, this study manipulates typing errors. The study includes typographical errors, because previous research demonstrated that typographical errors had more impact on credibility than orthographical errors (Cox et al., 2017). Therefore, the reviews with typing errors include various typographical errors (i.e. mechanical errors, such as mistyping) and the reviews without typing errors does not. Each condition is shown the same text, but the texts that are shown to the typing errors conditions, of course, contain typing errors. Thus, aside from the typographical errors, all review texts were identical. It is important to note that all reviews are written in a positive tone, as this study will not take into account the effect of the valence of the review. The main reason for this is that the conditions that would portray the studio aesthetic image with a negative review are not realistic, as the studio aesthetic image portrays a favorable image of the hotel. Therefore, these conditions would not be logical, as a customer who leaves a negative review would not attach a high quality image. Moreover, as mentioned before, it has been found that customers are less likely to attach pictures to their review if they had a negative experience (Trpkovski et al., 2018).
In order to ensure that the reviews are perceived as realistic, existing reviews were analyzed and elements of these reviews were used in the reviews that were written by the researcher. The reviews were presented like on the hotel booking site Booking.com. However, some adjustments to the presentation were made, as the name of the reviewer, the grade that was given and other personal information were removed from the review, in order to make sure that these characteristics did not influence the judgement of the participants. The review was written on the basis of a few requirements. The first requirement is that the review should not be too long. Previous research has found that positive reviews have a much shorter length and are less detailed than negative reviews, as customers use more words to express their
25 frustrations and anger (Zhao et al., 2019). Moreover, a study done by Cao et al. (2011) found that review helpfulness rapidly declines after a length of 144 words. Next to the length, the review should discuss at least two topics that are often discussed in hotel reviews. According to Mankad et al. (2016), the most discussed topics in hotel reviews are the location, amenities (such as breakfast, wifi, etc.) and experience, for example with regards to the staff. Lastly, the review has to match the attached image to some extent. Therefore, the review should for instance not mention a view that cannot be seen on the picture.
In order to test whether participants recognized the typing errors, a pre-test was conducted among 10 participants. The pre-test was qualitative in nature, meaning that interviews were conducted. During the interviews, participants were asked to look at two texts, with one including typing errors and the other not. Aside from the typing errors, the reviews were identical. The participants were asked to elaborate on the extent to which they view the reviews as realistic and what stands out to them. Moreover, participants were asked to compare the stimuli and point out the differences they recognize. Participants were also asked to think about what kind of person wrote the review and took the picture (all questions can be found in appendix A). The results of the pre-test also showed that the text with the typographical errors was not seen as realistic, as it contained too many typing errors. The review was therefore adjusted by reducing the amount of typing errors in the text. Figures 2 and 3 below are the adjusted reviews that were eventually used in the questionnaire.
26 Figure 2: Review without typographical errors
Figure 3: Review with typographical errors
27 In order to manipulate image style, two types of images were included. The image with the studio aesthetics is a professionally taken picture that can be found on the hotel’s website. Thus, an image of an existing hotel from the hotel’s website or from a booking platform was used. As stated by Ke et al. (2006), this professional picture is high in contrast, colorfulness and brightness and is sharp without noise. The image shows a clear and clean picture of a hotel bedroom, meaning no clutter such as personal belongings or an unmade bed is visible (see figure 4 below). The snapshot image is a picture that can be perceived as taken by a customer of the hotel. This picture was taken from the review section of a hotel booking website and corresponds with the studio aesthetics image. The sharpness and noise in this picture were not manipulated and are of rather high quality, as the devices that capture images nowadays are more advanced because of technological developments (Trpkovski et al., 2018).
However, the contrast, colorfulness and brightness is much lower, resulting in an overall lower quality image. Moreover, Ke et al. (2006) mentioned that snapshots often include more clutter. Therefore, the snapshot picture of the hotel bedroom shows some personal belongings as well as a not as neatly made bed (see figure 5 below for an example). Both images were taken in the daylight and consist of the same elements of the hotel room (both pictures show the whole bed for example), in order to allow for better comparison.
During the before mentioned pre-test, participants were also asked to look at four sets of images including one studio aesthetic image and one snapshot image from four different hotels in order to determine whether they recognized the different image styles (see appendix B). Participants were asked the same kinds of questioned as for the review texts, meaning they were asked what differences they recognized and what kind of person they thought the picture took. Moreover, the participants were asked to choose one set of images that represented the snapshots and studio aesthetics the best, according to them. The results of the pre-test showed that participants recognized the difference between the snapshot image and the studio
28 aesthetic image. Figure 4 and 5 below were found to best represent the studio aesthetic and the snapshot aesthetic according to 7 out of the 10 participants. This is therefore the set of images that was also used in the questionnaire. In appendix C an example can be found of how the stimuli for one condition were showcased in the questionnaire.
Figure 4: Studio aesthetic image of the hotel room 1
Figure 5: Snapshot image of the hotel room 1
29 3.3 Procedure
This section will describe the exact procedure that participants went through during the online experiment. When participants entered the questionnaire, they were first briefed about the purpose of the study, as well as how participating was entirely voluntarily and anonymous.
After the participants checked a box with which they gave consent to participate, participants got some demographical questions, such as their gender and educational level. Then they saw a short text in which they were asked to imagine that they were looking for a hotel, because they were planning a trip to Prague. Moreover, it was explained to the participant that information about the reviewer (such as their name), as well as the grade they gave, was removed from the review, because of privacy concerns. Participants were asked to carefully read the review and look at the picture (if they were given the condition that included a picture). Then each participant saw one of the six conditions, meaning they saw a review text with or without typing errors and without an image or with an image with snapshot or with studio aesthetics (see appendix C for an example). After looking at the review, the
participants were asked to what extent they agreed to various statements, first regarding purchase intention, then regarding review helpfulness, review trust, transaction trust and lastly authenticity. The survey ended with two questions regarding the extent to which participants use reviews to make purchase decisions and a manipulation check. The manipulation check asked two questions about whether the text contained typing errors, one questions to ask whether they saw an image and if they did, two questions about whether they saw a snapshot or a studio aesthetic image. Lastly, participants were thanked for their participation in the research.
3.4 Participants
A total amount of 238 participants were collected for this study. A convenience sample was used to recruit participants, meaning that the questionnaire was distributed via various
30 channels, among which e-mail, WhatsApp, LinkedIn, etc. Because everyone above 16 is a potential consumer of hotel experiences, no other limitations were set in place to filter
participants. As it was believed that this sampling technique would result in a wide variety of nationalities, the questionnaire was written in English.
After a first inspection of the dataset, it became clear that a significant amount of participants had left more than 4 questions blank. For this reason a total of 54 participants were removed from the dataset. Moreover, one more participant was deleted from the dataset, as they were underage. Therefore, a total amount of 183 participants were used for analysis.
The distribution of the participants among the conditions is equal, as each group contains about 30 participants. Table 2 below shows the demographic information of the respondents that were included in the dataset.
Table 2
Demographical information per condition
Note: M – mean value.
Group 1 – typing errors, studio aesthetics, Group 2 – no typing errors, studio aesthetics, Group 3 – typing errors, snapshots, Group 4 – no typing errors, snapshots, Group 5 – typing errors, no image, Group 6 – no typing errors, no image
Group1 Group 2 Group 3 Group 4 Group 5 Group 6
N % M N % M N % M N % M N % M N % M
Age 26.7 28.9 30.6 28.4 29.3 29.6
Gender Male 14 45.2 9 28.1 9 29 10 37 9 28.1 6 20
Female 17 54.8 22 68.8 22 71 17 63 23 71.9 24 80
Non binary 0 0 1 3.1 0 0 0 0 0 0 0 0
Level of education
Elementary school 1 3.2 0 0 2 6.5 0 0 0 0 0 0
High school 7 22.6 8 25 2 6.5 2 7.4 1 3.1 4 13.3
Vocational education 5 16.1 4 12.5 4 12.9 5 18.6 6 18.8 5 16.7
Bachelor’s degree in college 4 12.9 5 15.6 10 32.3 6 22.2 7 21.9 3 10
Bachelor’s degree in university
8 25.8 11 34.4 8 25.8 8 29.6 8 25 10 33.3
Master’s degree 6 19.4 4 12.5 3 9.7 6 22.2 10 31.3 8 26.7
Doctoral degree 0 0 0 0 2 6.5 0 0 0 0 0 0
Total 31 100 32 100 31 100 27 100 32 100 29 100
31 To test whether there were any significant differences between the groups for age, gender and educational level, three ANOVA tests were performed. The results of the test showed that age (F(5, 176) = 0.39, p = .858), as well as gender (F(5, 177) = 1.10, p = .361) and educational level (F(2, 177) = 1.21, p = .306) were not significantly different for each group. Therefore, it can be concluded that all groups have similar demographics.
3.5 Measures
Next to general demographical information (such as gender, age and experience with review sites), the dependent variables purchase intention, helpfulness transaction trust and review trust were measured.
In order to measure purchase intention a combination of the scale used by Bian and Forsythe (2012), who adopted it from Dodds, Monroe, and Grewal (1991), and the scale from Sharma et al. (2021) was used. A total of six items were included in the questionnaire to measure purchase intention (see appendix D for an overview of all items). A five-point Likert scale was used. An example of an item is: “If I were going to book a hotel, I would consider booking this hotel.”
To measure helpfulness both the scale by Sen and Lerman (2007) and the scale by Wu (2013) was used. Five items were included in order to measure the helpfulness of the review.
Again a five-point Likert scale was used. One of the items that was used is: “This review would aid my purchase decision.”
To measure both transaction trust and review trust a new scale was developed, which was inspired by the scales of Ohanian (1990), Bart et al. (2005) and Sparks and Browning (2011). Six items were constructed for review trust, and four items for transaction trust. All items used a five-point Likert scale. For review trust an item that was used is: “This review is sincere”. An example of an item that was used for transaction trust is: “I have confidence in purchasing this hotel experience.”
32 Next to the above mentioned scales that are part of the conceptual framework, the decision was made to measure the concept of authenticity. The choice was made to add an authenticity scale, because authenticity is highly related to review trust. As was briefly discussed in the theoretical framework, a study done by Kim and Kim (2020) showed that perceived authenticity of an online review has a significant impact on the development of trust. Therefore, by measuring authenticity, it might be the case that some of the effects that will be found during the statistical testing can be explained by authenticity. In order to determine whether the reviews were perceived as authentic, a scale by Banerjee and Chua (2021) was used. This scale consists of four items. All items were measured with a five-point Likert scale. An example of an item is: “This review is a genuine account of a post-trip experience”.
Lastly, a manipulation check was set in place, in order to determine whether the participant recognized the condition they were in. This check consists of five questions in total in which the participant was asked two questions about whether heir review contained typing errors and one question to ask whether their review contained an image. If they answered yes to the question whether review contained an image, they were asked two questions about whether they found the image professional looking or if it looked like it was taken by a customer.
3.6 Validity and reliability of the measurements
A confirmative factor analysis was conducted to establish the validity of the measurements.
Prior to the factor analysis, the suitability of the data for factor analysis was assessed. As the KMO value was above .06 (KMO = .95) and Bartlett’s Test of Sphericity was significant (X2 (276) = 3906.90, p < .001), it was shown that the data was suitable for a factor analysis. The factor analysis was performed with varimax rotation on a total of 25 components. A fixed number of 5 components were extracted, as existing scales were used in the study.
33 Not all items loaded cleanly into the constructs they were supposed to measure, as authenticity and review trust loaded into the same component. This makes sense, as the constructs are highly related. However, the decision was made to not merge them into one component, as authenticity and review trust were used for different analysis later on.
Moreover, one of the items from transaction trust (number 4) loaded very low. A possible explanation for this is that the item was the only reversed item in the scale, meaning that participants possibly looked over the questions and answered them the same as the others.
Although the item was reverse coded, it loaded very low, so the decision was made to remove the item from the scale. After removing the item and performing the factor analysis again, all items loaded above .5, meaning that all factors are strong and form one component, therefore confirming the validity of the constructs.
After all items were merged into their respective constructs, the Cronbach’s Alpha was calculated in order to assess the internal consistency of the constructs and therefore the
reliability of the measurements. The results showed that all values were satisfactory, as they are all above .7. It can therefore be stated that the constructs are internally consistent and reliable. Table 3 below shows the outcome of the factor analysis, as well as the Cronbach’s Alpha for each construct.
34 Table 3
Results of the second factor analysis with varimax rotation and reliability analysis
Constructs
Component
α 1 2 3 4 5
Purchase Intention 1
If I were going to book a hotel, I would consider booking this hotel.
.93 .70
2 I would recommend this hotel to my friends and family.
.52 .56
3 My willingness to book a room in this hotel would be high if I were shopping for a hotel.
.75
4 I would probably choose this hotel over other hotels in the same area.
.74
5 I predict that I would buy this hotel experience if I am looking for a hotel in the future.
.71
6 If I were going to book a hotel, I would consider booking this hotel.
.72
Helpfulness 1 I would consider this review to be useful. .90 .65 2 This review helps me to make a decision about
booking this hotel.
.78
3 I think this review is informative. .42 .64 4 I would use this review when making a decision
for a hotel.
.74
5 This review would aid my buying decision. .79
Review Trust 1 This review is reliable. .93 .67 .43
2 I believe this review to be trustworthy. .72 3 I think this review is dependable. .42 .61
4 This review is sincere. .74
5 I have confidence in this review. .63 .46
6 Overall, I think this review is believable. .63 .45 Transaction
Trust 1
I have confidence in buying this hotel experience. .89 .62
2 I would experience enough trust to book this hotel.
.60
3 If I would book this hotel, I believe it would deliver on its promises.
.49 .55
4 I would not feel confident engaging in a transaction to buy this hotel experience.
-.95*
Authenticity 1 The review is a genuine account of a post-trip experience.
.90 .81
2 The review is written after a stay in the hotel. .70 3 The review is an honest description of a stay in
the hotel.
.79 4 Overall, I think this review is authentic. .84
* Deleted item from the transaction trust scale
